2020-07-24T08:32:58Z
Acceptance in OA@INAF
A multiwavelength investigation of candidate millisecond pulsars in unassociated
þÿ³-ray sources
Title
Salvetti, D.; MIGNANI, Roberto; DE LUCA, Andrea; MARELLI, MARTINO;
Pallanca, C.; et al.
Authors
10.1093/mnras/stx1247
DOI
http://hdl.handle.net/20.500.12386/26605
Handle
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
Journal
470
Number
A multiwavelength investigation of candidate millisecond pulsars
in unassociated
γ -ray sources
D. Salvetti,
1‹R. P. Mignani,
1,2A. De Luca,
1,3M. Marelli,
1C. Pallanca,
4A. A. Breeveld,
5P. H¨usemann,
6A. Belfiore,
1W. Becker
6,7and J. Greiner
61INAF – Istituto di Astrofisica Spaziale e Fisica Cosmica Milano, via E. Bassini 15, I-20133 Milano, Italy 2Janusz Gil Institute of Astronomy, University of Zielona G´ora, Lubuska 2, PL-65-265 Zielona G´ora, Poland 3INFN – Istituto Nazionale di Fisica Nucleare, sezione di Pavia, via A. Bassi 6, I-27100 Pavia, Italy
4Dipartimento di Fisica e Astronomia, Universit`a degli Studi di Bologna, Viale Berti Pichat 6-2, I-40127 Bologna, Italy 5Mullard Space Science Laboratory, University College London, Holmbury St Mary, Dorking, Surrey RH5 6NT, UK 6Max-Planck Institut f¨ur extraterrestrische Physik, Giessenbachstrasse 1, D-85741 Garching, Germany
7Max-Planck Institut f¨ur Radioastronomie, Auf dem H¨ugel 69, D-53121 Bonn, Germany
Accepted 2017 May 18. Received 2017 May 12; in original form 2017 January 16
A B S T R A C T
About one-third of the 3033γ -ray sources in the Third Fermi-LAT Gamma-ray Source Cat-alogue (3FGL) are unidentified and do not have even a tentative association with a known object; hence, they are defined as unassociated. Among Galacticγ -ray sources, pulsars rep-resent the largest class, with over 200 identifications to date. About one-third of them are millisecond pulsars (MSPs) in binary systems. Therefore, it is plausible that a sizeable frac-tion of the unassociated Galacticγ -ray sources belong to this class. We collected X-ray and optical observations of the fields of 12 unassociated Fermi sources that have been classified as likely MSPs according to statistical classification techniques. To find observational support for the proposed classification, we looked for periodic modulations of the X-ray and optical flux of these sources, which could be associated with the orbital period of an MSP in a tight binary system. Four of the observed sources were identified as binary MSPs, or proposed as high-confidence candidates, while this work was in progress. For these sources, we present the results of our follow-up investigations, whereas for the others we present possible evidence of new MSP identifications. In particular, we discuss the case of 3FGL J0744.1−2523 that we proposed as a possible binary MSP based upon the preliminary detection of a 0.115 d periodicity in the flux of its candidate optical counterpart. We also found very marginal evi-dence of periodicity in the candidate optical counterpart to 3FGL J0802.3−5610, at a period of 0.4159 d, which needs to be confirmed by further observations.
Key words: stars: neutron – stars: variables: general.
1 I N T R O D U C T I O N
Observations carried out with the Fermi Gamma-ray Space
Tele-scope have produced an unprecedented harvest of γ -ray sources
thanks to the improved performance of its Large Area Telescope (LAT; Atwood et al.2009) with respect to its predecessor, the Ener-getic Gamma Ray Experiment Telescope (Thompson et al.1993), which flew on the Compton Gamma-ray Observatory (CGRO). The most recent catalogue ofγ -ray sources detected by the LAT, the Third Fermi-LAT Gamma-ray Source Catalogue (3FGL; Acero et al. 2015), based on four years of data (2008–2012), contains 3033 sources, a factor of 10 more than detected by the CGRO in
E-mail:salvetti@lambrate.inaf.it
a similar time span (1991–1995; Hartman et al.1999). The identi-fication of these sources, however, is in many cases a challenging task owing to the still relatively largeγ -ray error regions and to the problems of identifying a signature that can unambiguously reveal the source nature. In the case ofγ -ray pulsars, the detec-tion of pulsadetec-tions represents such a signature, and over 200γ -ray sources have been identified in this way1(see Grenier & Harding
2015for a recent review onγ -ray pulsars), which are either radio loud (RL), i.e. also detected as radio pulsars, or radio quiet (RQ), that is, undetected in radio in spite of deep searches. In the case of active galactic nuclei (AGNs), the second most numerous class of
1For the public list of γ -ray pulsars detected by the LAT, see
https://confluence.slac.stanford.edu/x/5Jl6Bg.
2017 The Authors
identifiedγ -ray sources, the characteristic signature is represented by the observation of long-term variability, possibly correlated with a similar activity in the radio band. The number of sources in the 3FGL that have been identified through a characteristic signature represents, however, a tiny minority. Indeed, while 1785 γ -ray sources in the 3FGL have been at least associated with objects in master catalogues based upon positional coincidence (mainly AGNs), another 1010 sources have no association at all, hence re-ferred to as unassociated. Needless to say, ascertaining that the nature of these sources has far reaching implications in the under-standing of theγ -ray emission of our Galaxy.
Sinceγ -ray pulsars represent the most numerous class of iden-tified Galacticγ -ray sources, it comes as natural that many efforts have been focused on the search for these objects. Apart from rela-tively young (1 Myr old) γ -ray pulsars, the LAT has identified a surprisingly large number of millisecond pulsars (MSPs). These are old (1 Gyr) pulsars, much less energetic than the younger ones, with a rotational energy loss rate of ˙Erot∼ 1033erg s−1, which owe
their fast rotation periods (a few milliseconds, hence the name) to the spin-up following a phase of matter accretion from a donor com-panion star. Not surprisingly, out of the 93γ -ray MSPs, 73 are found in binary systems. They have usually either a white dwarf (WD) or a low-mass main sequence (MS) companion, such as the so-called black widows (BWs) and redbacks (RBs), which have compan-ion masses Mcom< 0.1 M and Mcom∼ 0.1–0.4 M, respectively
(Roberts2013). The low companion masses explain the short or-bital periods (Porb < 1 d) of these systems and fit the scenario
where the companions are ablated by irradiation from the pulsar wind, a process that would explain the existence of solitary MSPs. This scenario has recently received observational support from the identification of some RB systems that alternates between states of accretion from the companion, when the X-ray emission increases and the radio emission is temporarily quenched by the accreting material, and states of no accretion, where the X-ray emission de-creases and the radio emission reactivates (e.g. Patruno et al.2014), hence dubbed ‘transitional’.
Many BWs and RBs have been found to lurk in unassociated
γ -ray sources detected by Fermi, and their number has now
in-creased by at least a factor of 5. The difficulties in finding such systems in the traditional radio channel, that is, by detecting them as radio MSPs, is intrinsic to the very nature of such systems. Even in the non-accretion phases, a considerable amount of plasma is present in the intrabinary environment, produced by the irradiation of the companion star surface, which causes eclipses and delays of the radio emission from the MSP. This makes BWs and RBs very elusive targets in radio and their identification requires ob-servations at different wavelengths, e.g. in the X-rays and in the optical. In a number of cases, these observations have been instru-mental in pinpointing candidate BW/RB systems in unassociated
Fermi sources and paved the way for the discovery of radio/γ -ray
pulsations. Since BWs and RBs are characterized by short orbital periods, the most successful strategy consists of searching for<1 d periodic modulations of the optical flux of the putative companion star. These are produced by irradiation and tidal distortion effects that affect the star brightness when it is seen at different orbital phases. This strategy was applied to pinpoint the MSPs associ-ated with theγ -ray sources 1FGL J1311.7−3429 (Romani et al.
2012) and 1FGL J2339.7−0531 (Kong et al.2012), and other likely MSP candidates associated with 2FGL J1653.6−0159 (Romani, Filippenko & Cenko2014), 2FGL J0523.3−2530 (Strader et al.
2014), 3FGL J2039.6−5618 (Romani2015; Salvetti et al.2015) and 3FGL J0212.1+5320 (Li et al.2016).
How many unassociatedγ -ray sources, especially at high Galac-tic latitude where MSPs have time to migrate on Gyr time-scales thanks to their proper motions, can be identified as either BWs or RBs, is an open issue. To help addressing this issue, back in 2014 we started a project to search for BWs/RBs in a selected sample of unassociated Fermi-LAT sources based on the detection of op-tical modulations with a few hour periods from the putative MSP companions. A preliminary account of our project, with the discus-sion of early results for the most interesting sources, is given in Mignani et al. (2016). Our manuscript is structured as follows: in Section 2, we describe the selection of the MSP candidates and the multiwavelength observations. In Section 3, we present the results, and discuss their implications in Section 4. Finally, we conclude in Section 5.
2 O B S E RVAT I O N S A N D DATA R E D U C T I O N 2.1 Candidate selection
As a first step, we selected a starting sample of MSP candidates from the unassociated Fermi-LAT sources. Since our project started in 2014, we originally selected these sources from the 2FGL catalogue (Nolan et al.2012), which was the reference catalogue of Fermi-LAT sources available back then. All the sources selected are also listed in the 3FGL catalogue (Acero et al.2015), which from now on we assume as a reference. The candidate selection was based on their
γ -ray spectral and temporal characteristics and was implemented
through the results of an artificial neural network classification code, fully described in Salvetti (2016). Our candidate selection method agrees with that of Saz Parkinson et al. (2016), which also classified all our candidates as MSPs. This starting sample was the basis of our multiwavelength investigations aimed at confirming the source classification on firm observational evidence.
As a second step, we narrowed down the candidate selection to those γ -ray sources with small γ -ray error ellipses (r95 ≤ 0.◦1) and with X-ray coverage from either XMM–Newton, Chandra or
Swift. This is because MSPs are also X-ray sources, with the X-ray
emission produced from the pulsar magnetosphere (or hot polar caps on the pulsar surface) and/or from the intrabinary shock (Roberts
2013).
Finally, we selected candidates for which multi-epoch photom-etry measurements were available from public optical sky surveys and/or for which we obtained follow-up imaging observations with ground-based optical facilities (see Section 2.3 for details). In this way, we selected a sample of 12 MSP candidates, including the suspected RB candidates 3FGL J0523.3−2528, J1653.6−0158 and J2039.6−5618 (Table1). We refer to Acero et al. (2015) for the characterization of theγ -ray properties of these sources and to Saz Parkinson et al. (2016) for the description of their classification as MSP candidates. An account of the available X-ray and optical observations is given in the next two sections and is summarized in Table1.
2.2 X-ray observations
All our targets have an adequate X-ray coverage of the LAT er-ror boxes with either Swift (Burrows et al.2005) or XMM–Newton (Str¨uder et al.2001; Turner et al.2001). All of them have been ob-served by Swift as part of a systematic survey of theγ -ray error boxes of the unassociated Fermi-LAT sources (Stroh & Falcone 2013). In all cases, the XMM–Newton observations have been executed as pointed follow-ups of the LAT sources. Only 3FGL J1653.6−0158
Table 1. Candidate MSPs from unassociated Fermi-LAT sources discussed in this work. Coordinates and size of the 95 per cent semimajor axis of theγ -ray error ellipse (r95) are taken from the 3FGL catalogue (Acero et al.2015). Sources are sorted in RA. The multiwavelength observations used in this work are summarized in columns five and six.
3FGL name RA DEC r95 X-ray observations Optical observations Notes (references)
(hh mm ss) (◦ ) (◦)
J0523.3−2528 05 23 21.4 −25 28 35 0.04 Swifta Catalinaa, GROND Candidate RB (1)
J0744.1−2523 07 44 10.7 −25 23 58 0.05 Swift GROND
J0802.3−5610 08 02 19.9 −56 10 08 0.10 Swift, XMM–Newton Catalina
J1035.7−6720 10 35 42.2 −67 20 01 0.04 Swift, XMM–Newton GROND Pulsar (2)
J1119.9−2204 11 19 56.3 −22 04 02 0.04 Swiftb, XMM–Newton Catalinab, GROND
J1539.2−3324 15 39 17.6 −33 24 51 0.04 Swiftb Catalina, GROND
J1625.1−0021 16 25 07.1 −00 21 31 0.04 Swift, XMM–Newtonb Catalina, GROND
J1630.2+3733 16 30 12.8 +37 33 44 0.07 Swift Catalina Binary MSP (3)
J1653.6−0158 16 53 40.6 −01 58 48 0.04 Swift, Chandrab Catalinaa Candidate RB (4)
J1744.1−7619 17 44 10.8 −76 19 43 0.03 Swift, XMM–Newtonb Catalina Pulsar (2)
J2039.6−5618 20 39 40.3 −56 18 44 0.04 Swiftc, XMM–Newtonc Catalinac, GRONDc Candidate RB (5,6)
J2112.5−3044 21 12 34.7 −30 44 04 0.04 Swift, XMM–Newton Catalina
Notes. List of references: (1) Strader et al. (2014); (2) Clark et al. (2017); (3) Sanpa-Arsa et al. (in preparation); (4) Romani et al. (2014); (5) Salvetti et al. (2015) and (6) Romani (2015).
Data analysis published inaStrader et al. (2014);bHui et al. (2015);cSalvetti et al. (2015) and Romani (2015).
(Cheung et al. 2012) has been observed with Chandra (Garmire et al.2003).
We reduced and analysed the XMM–Newton data through the most recent release of the XMM–Newton Science Analysis Soft-ware (SAS) v15.0. We performed a standard data processing, using
the epproc and emproc tools, and screening for high particle background time interval following Salvetti et al. (2015). For the
Chandra data analysis, we used the Chandra Interactive Analysis of
Observation (CIAO) software version 4.8. We recalibrated event data
by using the chandra_repro tool. Swift data were processed and filtered with standard procedures and quality cuts2using FTOOLS
tasks in theHEASOFTsoftware package v6.19 and the calibration files
in the latest Calibration Database release.
2.3 Optical observations
Ten sources in our sample were observed as part of the Catalina Sky Survey3, a programme of three optical sky surveys covering
the whole sky 15◦ above and below the Galactic plane (Drake et al. 2009). In the Northern hemisphere, the Catalina Sky Sur-vey is carried out with two wide-field telescopes: the 0.7 m Schmidt at the Mount Bigelow Observatory (Arizona), which produces the Catalina Schmidt Survey (CSS), and the 1.5 m reflector at Mount Lemmon Observatory (Arizona), which produces the Mount Lem-mon Survey (MLS). In the Southern hemisphere, the 0.5-m Schmidt telescope (Siding Spring, Australia) was retired at the end of 2013 and only archival observations produced by the Siding Spring Sur-vey (SSS) are available. The Catalina Sky SurSur-vey is carried out with a cadence of days to months down to limiting magnitudes
V∼ 19.5–21.5 per pass.
We carried out dedicated follow-up observations of seven of the MSP candidates in our sample with the Gamma-Ray Burst Optical/Near-Infrared Detector (GROND; Greiner et al.2008) at the MPI/ESO 2.2-m telescope on La Silla (Chile). For five of these candidates, we have also multi-epoch data from the Catalina Sky Survey. We carried out simultaneous observations in the optical g,
r, i, zand near-infrared (IR) J, H, K bands repeated with a regular
2More detail in:http://swift.gsfc.nasa.gov/docs/swift/analysis/. 3http://www.lpl.arizona.edu/css/
cadence on two or more consecutive nights. The observations were split into sequences of four 115 s dithered exposures in the optical and 48 10 s dithered exposures in the near-IR. To cope with the right ascension distribution of our targets, the observations were divided in two runs, in 2014 August and 2015 February. The data were processed and calibrated using the GROND pipeline (Kr¨uhler et al.
2008; Yoldas et al.2008). The astrometry calibration was calculated from single exposures by comparison with stars selected from the USNO-B1.0 catalogue (Monet et al.2003) in the optical bands and the 2MASS catalogue (Skrutskie et al.2006) in the near-IR bands, yielding an accuracy of 0.3 arcsec with respect to the chosen refer-ence frame. The photometric calibration in the optical was computed using close-by fields from the Sloan Digital Sky Survey (York et al.
2000), whereas in the near-IR it was computed using 2MASS stars in the GROND field of view. The accuracy of the absolute photome-try calibration was 0.02 mag in the g, r, i, zbands, 0.03 mag in the
J and H bands and 0.05 in the Ksband. See H¨usemann (2015) for
details on the GROND data processing and calibration. A separate automated variability analysis was then executed on the photome-try result files of the standard GROND pipeline for all sources in the field of view using a code developed at Max-Planck Institut f¨ur extraterrestrische Physik, based on a standard deviation analysis of the optical light curve (H¨usemann2015) and with a threshold
σth= 0.05. In parallel, we ran an independent periodicity search
using the Lomb–Scargle algorithm (Zechmeister & K¨urster2009), exploring candidate periods above 0.1 d, to include the period range characteristic of BW and RB systems. We then inspected candidate periodicities through a standard folded light-curve analysis.
2.4 X-ray/optical analysis
As a first step, we looked for candidate X-ray counterparts to the LAT sources that had not yet been firmly identified in the X-rays, that is, all sources in Table1except the three RB candidates. We performed a standard data analysis and source detection in the 0.3– 10 keV energy band of the XMM–Newton-EPIC, Chandra-ACIS and Swift-XRT observations (e.g. Marelli et al.2015; Salvetti et al.
2015). We focused on the X-ray sources detected at a significance 3σ inside, or close to, the 95 per cent confidence 3FGL error ellipse to search for the possible counterparts to eachγ -ray source.
For each of these X-ray sources, we performed a spectral and tim-ing analysis ustim-ingXSPECv12.9 andXRONOSv5.21, respectively. We
extracted X-ray fluxes by fitting the spectra with a power-law (PL) model using either aχ2or the C-statistic (Cash1979) in the case of
low counts (<100 photons). For sources characterized by low statis-tics, we fixed the column density to the value of the Galactic NH
integrated along the line of sight (Dickley & Lockman1990) and, if necessary, the X-ray PL photon index (X) to 2. All quoted
uncer-tainties on the spectral parameters are reported at the 1σ confidence level. For all sources, we computed the correspondingγ -to-X-ray flux ratio. As reported in Marelli et al. (2015), this could still give important information on the nature of the source. For all sources, we also generated background-subtracted X-ray light curves with at least 25 counts per time bin in order to evaluate the variability sig-nificance through aχ2test. Since BWs and RBs are characterized
by an X-ray flux modulation with a period smaller than 1 d, which is associated with emission from the intrabinary shocks (Salvetti et al.2015), we searched for periodic modulations in the barycen-tred data using the standard power spectrum analysis. Finally, for all observations, we computed the 3σ X-ray detection limit based on the measured signal-to-noise ratio, assuming a PL spectrum with
X= 2 and the integrated Galactic NH.
As a second step, we looked for variable optical counterparts to the detected X-ray sources in either the Catalina Sky Survey or in the GROND data (or both). Therefore, our strategy is tailored to the identification of binary MSPs. Only when we found a potentially interesting counterpart, that is, with clear evidence of periodic flux modulation, we also exploited multiband observations in optical, ultraviolet and IR archives for a better counterpart characterization. We note that it was not possible to use the Catalina Surveys Periodic Variable Star Catalogue (Drake et al.2014) to carry out a systematic search for variable sources across the entire 3FGL error ellipses of ourγ -ray sources since this catalogue only covers the declination region−22◦< δ < 65◦and all our targets, with the only exception of 3FGL J1625.1−0021, J1630.2+3733 and J1653.6−0158, are south of it.
3 R E S U LT S
Out of the 10 LAT sources with optical coverage in the Catalina Sky Survey (Table 1), we could recover the periodic-ity of the optical counterparts to the two known RB candidates 3FGL J0523.3−2528 and 3FGL J1653.6−0158, while for the third one (3FGL J2039.6−5618) we could not find any clear evidence of periodicity in the Catalina data, despite finding evidence in the GROND data (see the next paragraph). For the other seven LAT sources, we could not find either an X-ray source in the 3FGL er-ror box (3FGL J1539.2−3324 and J1630.2+3733), or we found X-ray sources with no Catalina counterpart (3FGL J1744.1−7619 and J2112.5−3044), or X-ray sources that do have a Catalina coun-terpart but that is not periodically variable (3FGL J1119.9−2204 and J1625.1−0021). For 3FGL J0802.3−5610, we found an X-ray source associated with a Catalina counterpart with only a marginal evidence of periodicity that requires to be confirmed by further optical observations. Out of the seven sources that have also (or only) GROND coverage, we could detect a clear period-icity for the optical counterpart to 3FGL J2039.6−5618 (Salvetti et al.2015). For 3FGL J0523.3−2528, we could only partially
re-cover the light curve of its optical counterpart, obtained from the Catalina data, owing to a sparse phase coverage of the GROND data. Like 3FGL J1539.2−3324 (see above), 3FGL J0744.1−2523 also has no candidate X-ray counterpart; however, despite this, we
found a candidate GROND counterpart in the 3FGL error ellipse with clearly periodic modulation. Both 3FGL J1119.9−2204 and J1625.1−0021 have candidate X-ray counterparts with potential GROND counterparts, but they are not periodically variable. Fi-nally, 3FGL J1035.7−6720 has candidate X-ray counterparts but it has no potential GROND counterparts.
One of the sources in our sample has been identified as a bi-nary MSP (3FGL J1630.2+3733; Sanpa-Arsa et al. in prepara-tion). In addition, while this work was being finalized, another two (3FGL J1035.7−6720 and 3FGL J1744.1−7619) were detected as pulsars by Clark et al. (2017), although no value of the spin (and or-bital) period has been published, leaving unspecified whether they are MSPs or not and whether they are binary or isolated. None of the sources in our sample has a radio association in the recent follow-up by Schinzel et al. (2017). In the following, we describe the results on a case-by-case basis.
3.1 High-confidence binary MSP candidates
3.1.1 3FGL J0523.3−2528
An interesting X-ray candidate counterpart to this originally unas-sociatedγ -ray source was found by Acero et al. (2013) in a 4.8 ks Swift observation (see also Takeuchi et al. 2013). This is the only X-ray observation available for this source. Soon after, based on an analysis of the Catalina V-band data and on follow-up ra-dial velocity spectroscopy observations, Strader et al. (2014) found a periodically modulated optical counterpart (0.688 d), with the light curve featuring two peaks. This sets the case for the identi-fication of 3FGL J0523.3−2528 as a new binary MSP candidate, possibly an RB. In the early phases of our project, we indepen-dently found the optical periodicity in this source using the same Catalina data set as used in Strader et al. (2014). This triggered a proposal for follow-up GROND observations submitted for the 2014 September/2015 March semester. The GROND observations were eventually performed in 2015 February. Unfortunately, the non-optimal scheduling of the requested observation sequence did not allow us to uniformly cover in phase the 0.688 d cycle, with most measurements covering about half of the two peaks in the light curve.
Fitting the X-ray spectrum with a PL model withX= 1.63 ± 0.2
and NHfixed to the integrated Galactic value of 1.9× 1020cm−2,
we obtain an unabsorbed 0.3–10 keV flux FX = (1.85 ±
0.3)× 10−13 erg cm−2s−1. For aγ -ray flux above 100 MeV of
Fγ = (1.99 ± 0.12) × 10−11erg cm−2s−1(Acero et al.2015), the
γ -to-X flux ratio is Fγ/FX∼ 100.
3.1.2 3FGL J1653.6−0158
Candidate X-ray counterparts to 3FGL J1653.6−0158 were found by Cheung et al. (2012) using a Chandra-ACIS 21 ks observation (OBSID 11787). An optical follow-up of the 3FGL J1653.6−0158 field by Romani et al. (2014) led to the discovery of a periodic flux modulation (0.052 d) in the optical counterpart to the brightest of the Chandra sources, making a case for a new binary MSP identified through the detection of an optical periodicity, also in this case a likely RB. The periodicity was also found in the Catalina data, allow-ing Romani et al. (2014) to extend the time baseline for the period determination. We note that the source (MLS J165338.1−015836) is not included in the Catalina Surveys Periodic Variable Star Cat-alogue (Drake et al.2014), since this includes data from the CSS only and not from the MLS. The detection of periodicity in the
optical counterpart to the X-ray source was confirmed by Kong et al. (2014) from the analysis of independent observations. The spec-trum of the X-ray source was fitted with a PL withX= 1.65+0.39−0.34
(NH= 1.3+1.8−1.3× 1021 cm−2) in the 0.5–8 keV energy range
(Romani et al.2014). These values yield an unabsorbed 0.3–10 keV X-ray fluxFX= 2.3+0.9−0.6× 10−13erg cm−2s−1. Itsγ -ray flux above
100 MeV is Fγ= (3.37 ± 0.18) × 10−11erg cm−2s−1(Acero et al.
2015), giving an Fγ/FX∼ 150.
More recently, the Chandra data have been reanalysed by Hui et al. (2015), who claimed possible evidence (at the 99.2 per cent confidence level) of an X-ray modulation in this source, at the same 0.052 d period as observed in the optical. No evident X-ray modu-lation, however, was found by Romani et al. (2014), whereas only a tentative evidence was found by Kong et al. (2014). We reanal-ysed the same Chandra data to verify the existence of a possible X-ray modulation. However, our analysis of the folded 0.3–10 keV light curve does not show any evidence of deviation from a con-stant flux at a level above 3σ . Therefore, our conclusions are in line with those of Romani et al. (2014) and Kong et al. (2014). Like for 3FGL J0523.3−2528, also for 3FGL J1653.6−0158 we indepen-dently found in the Catalina data the same periodicity as discovered by Romani et al. (2014).
3.1.3 3FGL J2039.6−5618
As a part of this project, we observed the unassociated Fermi-LAT source 3FGL J2039.6−5618 with both XMM–Newton (OBSID 0720750301) and GROND and we identified it as a likely RB from the discovery of a common periodicity in its X-ray and optical/near-IR counterpart at a period of 0.227 d (Salvetti et al.2015), which we identified as the orbital period of a tight binary system, where one of the members is a neutron star.
We also detected the optical counterpart to the X-ray source in the Catalina Sky Survey (217 epochs), but the large error bars at-tached to the single flux measurements did not allow us to confirm the periodicity observed in the GROND data and search for pos-sible long-term evolution of the light curve. The periodicity has been independently discovered by an analysis of our GROND data by Romani (2015), who complemented them with data taken with the Southern Astrophysical Research and Dark Energy Survey tele-scopes, allowing him to extend the time baseline for the light-curve folding and improve on the determination of the period accuracy. The X-ray and optical light curves are shown in Fig.1(top and middle panel, respectively), aligned in phase using the updated pe-riod determined by Romani (2015), PB= 0.228 116 ± 0.000 002 d,
and the epoch of quadrature (MJD= 56884.9667 ± 0.0003) deter-mined by Salvetti et al. (2015) by fitting the GROND light-curve profile with a geometrical model of the tidally distorted neutron star companion (see also Mignani et al.2016).
In Salvetti et al. (2015), we used the GROND data to characterize the phase-averaged spectrum of the X-ray source counterpart to-gether with optical/ultraviolet data from the XMM–Newton Optical Monitor (OM; Mason et al.2001) and the Swift Ultraviolet Optical Telescope (UVOT; Roming et al.2005). To these data, we added mid-IR flux measurements from the archival Wide-field Infrared
Survey Explorer (WISE; Wright et al.2010) data, obtained in the W1 (3.4µm), W2 (4.6 µm), W3 (12 µm), W4 (22 µm) bands, and avail-able in the AllWISE catalogue (Wright et al.2010). The multiband UV-to-mid-IR spectrum is shown in Fig.1(bottom). Interestingly, as qualitatively shown in Mignani et al. (2016), the new WISE data confirm that the source spectrum features a cold blackbody (BB) with effective temperature Teff= 1700 ± 120 K, dominating in the
0 0.5 1 1.5 2
0.5
1
1.5
NORMALIZED FOLDED LIGHT CURVE
PHASE
XMM−Newton EPIC
Figure 1. Top to bottom: XMM–Newton light curve of the 3FGL J2039.6−5618 counterpart, GROND light curve and multiband spectrum of its optical counterpart from GROND, UVOT, OM and WISE data. The X-ray and optical light curves are aligned in phase. The black dotted lines in the bottom panel are the two BB components to the spectrum (blue dashed line) best fitting the data points (red). Filters are labelled with their names.
near/mid-IR, and a hot BB, with Teff= 3800 ± 150 K, dominating
in the optical/ultraviolet. We looked for a periodic flux modula-tion in the multi-epoch WISE data but we found no evidence of it, within the statistical uncertainty of the WISE flux measurements (∼0.2–0.5 mag in the W1 band). As we anticipated in Mignani et al. (2016), the lack of a modulation at the optical/near-IR period would suggest that the mid-IR emission component of the spectrum is not
produced by the tidally distorted companion star but by a different source in the binary system.
The spectrum of the 3FGL J2039.6−5618 X-ray counterpart was fitted with a PL (X= 1.36 ± 0.09; NH< 4 × 1020cm−2), with
an unabsorbed 0.3–10 keV X-ray fluxFX= 10.19+0.87−0.82× 10−14erg
cm−2 s−1 (Salvetti et al. 2015), which gives Fγ/FX ∼ 170 for
Fγ= (1.71 ± 0.14) × 10−11erg cm−2s−1(Acero et al.2015). 3.2 Possible binary MSP candidates
3.2.1 3FGL J0744.1−2523
The field of 3FGL J0744.1−2523 is not covered by the Catalina Sky Survey. However, we found an interesting GROND source within the 3FGL error ellipse (time-averaged magnitude r ∼ 19.18), at
α = 07h44m08.s47 andδ = −25◦2358.9 (J2000), which is clearly
variable (H¨usemann2015). This source features a clear flux modu-lation with an optical period equal to 0.115 42± 0.000 05 d and an amplitude of∼0.8 mag (Fig.3, top). We estimated the period uncer-tainty following Gilliland & Baliunas (1987). The significance of this modulation is 7.4σ , as computed from the generalized Lomb– Scargle periodogram algorithm (Zechmeister & K¨urster2009). We recognized the same modulation in the near-IR bands, where the best period is found at 0.115 46± 0.000 09 d. Therefore, this source could be a candidate companion to an MSP in a tight binary system. The field of 3FGL J0744.1 − 2523 was covered with a series of
Swif t observations (23 ks total integration). However, we did not
detect any associated X-ray source (Fig.2) down to a 0.3–10 keV flux limit FX= 4.5 × 10−14erg cm−2s−1(3σ level). For a γ -ray
flux Fγ= (2.38 ± 0.17) × 10−11erg cm−2s−1(Acero et al.2015), this would correspond to aγ -to-X-ray flux ratio Fγ/FX 525 for
3FGL J0744.1−2523. No other X-ray observations of this field have ever been performed.
Interestingly, the spectrum of the GROND source (Fig. 3, bottom) is very similar to that of the optical counterpart to 3FGL J2039.6−5618 (Fig.1, bottom), with cold and hot BBs at comparable temperatures, possibly suggesting similar characteris-tics for the companion star, as suggested in Mignani et al. (2016). Like we did for 3FGL J2039.6−5618, to investigate this possible similarity we searched for a mid-IR counterpart in the archival WISE data and found that the fluxes reported in the AllWISE catalogue (Wright et al.2010) are well above the composite spectrum best fitting the GROND data. However, inspection of the WISE images shows that the mid-IR source (Fig.4, left) is likely a blending of three (or more) sources that are clearly resolved in the higher spatial resolution GROND data (Fig.4, right). Therefore, the fluxes of the mid-IR source matched in the WISE data cannot be used to constrain the spectrum of the GROND source.
To complement the data presented in Mignani et al. (2016) and better define the source spectrum at shorter wavelengths, we looked for photometry data in the Swift-UVOT observations. There are four sets of observations taken in the V, B and U bands, the first (obsid 00041337001) taken in 2010 August, had only the U-band data. Three more observations (OBSIDs 00031960001, 2 and 3) were taken in 2011 April with all the three optical filters. We used the standard uvotmaghist and uvotsource FTOOLs with a 3 arcsec aperture at the source position, as computed from the GROND images, and an annulus between 11.6 and 16.8 arcsec from the source position for the background (smaller than the standard annulus because of the crowded field), to measure the flux of the source. This was clearly detected in a few hundred seconds of exposure time in both the V and B bands. We could not find any
evidence of significant variability across single exposures, so that we decided to co-add all of them to achieve higher signal-to-noise detections, with a corresponding integration time of 3527 and 9032 s in the B and V bands, respectively. In the U band, we could not detect the source in the single exposures. However, the source is detected in the exposure co-addition (10 296 s integration time), although at a marginal signal-to-noise∼3σ . The results of our photometry are
U= 23.12 ± 0.32, B = 21.07 ± 0.16, V=19.85 ± 0.09, where all
magnitudes have been converted into the AB system to be directly comparable with the GROND ones.
We used the multiband flux information on the GROND source to determine its nature from its colours. Fig.5shows the source position in the gversus g− rcolour–magnitude diagram (CMD), compared to that of the field sources (blue points) and stars simu-lated with the Besanc¸on models (Robin et al.2004) for different dis-tance values, that is, 1–3 kpc (light cyan) and<10 kpc (dark cyan). The grey points correspond to the variable GROND source, with the grey-scale varying with the phase of the light curve, that is, light and dark grey correspond to the phases of the light curve minimum and maximum, respectively. The CMD clearly shows that the posi-tion in the diagram fluctuates as a funcposi-tion of the colour evoluposi-tion along the light curve but remains along the sequence of field stars and is consistent with a late MS star at a distance of 1–3 kpc. Note that we neglected the effect of the extinction in the CMD. Such an assumption is justified by the properties of the colour–colour diagram for stars in the GROND field of view. In fact, we com-pared the observed colour–colour diagram (distance independent) with the expected colour–colour diagram based on the simulations with the Besanc¸on models and we found good agreement between the two without introducing any extinction value. The Galactic hy-drogen column density in the source direction is NH∼ 6 × 1021cm−2
(Dickey & Lockman1990), corresponding to E(B− V) ∼ 1.1 (Pre-dehl & Schmitt 1995). This value is larger than the modest ex-tinction values that we inferred from the colour–colour diagrams, which independently confirms that the star must be relatively close, as suggested by the comparison with the Besanc¸on models. For instance, a simple scaling of the NHto produce an interstellar
red-dening E(B− V) ∼ 0.11, that is, one-tenth of the integrated Galactic value, would put the source at a distance of≈1.5 kpc, perfectly con-sistent with the distance range inferred from the Besanc¸on models. We note that this object appears in the first Gaia data release,4but,
unfortunately, it has no measured parallax or proper motion yet.
3.2.2 3FGL J0802.3−5610
The field of 3FGL J0802.3−5610 has been observed by both Swift (8 ks) and XMM–Newton (18 ks). From the Swift data, we detected no X-ray source at a level of significance above 3σ in the 3FGL error ellipse of theγ -ray source. Therefore, we used the deeper XMM–
Newton observation (OBSID 0691980301) to search for a candidate
X-ray counterpart to 3FGL J0802.3−5610. For our X-ray analysis, we selected only 0–4 pattern events from the pn and 0–12 events from the two MOS detectors with the default flag mask. We ran the source detection in the 0.3–10 keV energy range simultaneously on the event lists of each of the EPIC-pn and MOS detectors using a maximum likelihood fitting with theSAStask edetect_chain
invoking other SAS tools to produce background, sensitivity and
vignetting-corrected exposure maps. The final source list includes 10 X-ray sources close to the 3FGL error ellipse detected from both
4http://www.cosmos.esa.int/web/gaia/dr1
Figure 2. 0.3–10 keV exposure-corrected X-ray images (15 arcmin× 15 arcmin) of the binary MSP candidates discussed in Section 3.2 obtained with the XMM–Newton-EPIC and Swift-XRT (3FGL J0744.1−2523 and 3FGL J1539.2−3324) instruments. All images have been smoothed using a Gaussian filter with a kernel radius of 3 arcsec. Images include the 95 per cent confidence level Fermi-LAT position error ellipse (shown in yellow), as given in the 3FGL catalogue (Acero et al.2015). Plausible X-ray counterparts to theγ -ray source, detected within or close to the LAT error ellipse, are highlighted with a white circle and labelled as in Table2, while other X-ray sources detected in the FoV are plotted with a red circle.
Figure 3. Top to bottom: GROND light curve and spectrum of the 3FGL J0744.1−2523 candidate counterpart. In the bottom panel, the two BB components to the best-fitting spectrum (blue dashed line) are shown as black dotted lines. Observed fluxes are shown by the red points. Filters are labelled with their names (updated from Mignani et al.2016). No interstellar extinction correction has been applied. All magnitudes are in the AB system.
the pn and MOS detectors, with a combined pn+MOS detection likelihood greater than 10, corresponding to a 3.5σ detection significance.
The XMM–Newton image of the 3FGL J0802.3−5610 field with the detected X-ray sources is shown in Fig. 2. The X-ray source properties are summarized in Table2. We detected a Catalina candidate counterpart only for Source #5. This is the brightest X-ray source detected within/close to the 3FGL error ellipse, with an unabsorbed flux in the 0.3–10 keV energy range of
FX= 9.6+5.7−3.6× 10−14erg cm−2s−1. The spectrum is very soft and,
using the C-statistics, we could fit it with a PL with photon index
X= 4.08+0.74−0.68and NHfixed to 1.8× 1021cm−2. If Source #5 were
the X-ray counterpart to 3FGL J0802.3−5610, this would have an
Fγ/FX of∼135, for Fγ = (1.30 ± 0.12) × 10−11 erg cm−2 s−1
(Acero et al.2015). Although the PL model provides a statistically acceptable fit (C= 5.65 with 2 degrees of freedom), the photon index is unrealistically large. By fitting the spectrum with an absorbed BB model with fixed NH, we obtained a temperature of
kT= 0.15 ± 0.03 keV with C = 3.93. The thermal scenario would
provide a more reasonable description of the X-ray emission of
Source #5.
The Catalina counterpart (SSS J080225.1−560543) to Source #5 is at an angular distance smaller than 1 arcsec (α = 08h02m25s.08
andδ = −56◦0543.8, J2000) from the best-fitting X-ray source position, compatible with the uncertainty on the XMM–Newton as-trometry. The source is quite bright, with a magnitude V ∼ 14.03 averaged over 218 epochs extending over a time span of about eight years. By running the generalized Lomb–Scargle periodogram algo-rithm on the light curve, we found that the most significant candidate period is at∼1 d with its harmonics. Examining the power spectrum of the window function, we verified that these periods are associ-ated with the cadence of the observations. Except for these periods, we detected another distinct peak located at a period of∼0.4159 d, characterized by a significance of∼5.3σ , after accounting for the number of trials. In addition, we observed several aliases of the candidate frequency at consistent offsets from each harmonic of the daily periodicity. Since the strong daily periodicity affects the statistics in the spectrum, we filtered the time series subtracting the daily sinusoid from the data as suggested by Horne & Baliunas (1986) to determine whether the tentative period is real. After fil-tering, we observed that the putative period becomes the strongest peak in the new periodogram but with a significance just below 5σ . It is clear that the 1-d signal in the observing window hampers a robust estimate of the statistical significance of the candidate sig-nal. Thus, we cannot draw firm conclusions about the reality of the∼0.4159 d modulation. Fig.6shows the Catalina light curve folded at the period of∼0.4159 d, which seems to feature a sin-gle, possibly asymmetric, peak. The field of 3FGL J0802.3−5610 has not been observed by GROND, so that we cannot confirm the putative periodicity seen in the Catalina data. Because of the poor counts statistic for Source #5 (∼40 net counts), we could not search for an X-ray flux modulation at the Catalina period (nor any other significant type of variation) in the XMM–Newton data.
Owing to the association with a potentially periodic vari-able optical counterpart, Source #5 is a promising counterpart to 3FGL J0802.3−5610. However, the lack of information about the optical spectrum of its candidate counterpart, together with the un-certain value of the best-fitting period and its soft, likely thermal, X-ray spectrum prevent us from discarding other plausible scenar-ios, such as a W UMa or aβ Lyr binary system or a cataclysmic variable (CV). Deeper X-ray observations and a better spectral and timing characterization of the optical counterpart are needed to de-termine the nature of the X-ray source.
Other X-ray sources detected within/close to the γ -ray error ellipse are fainter than Source #5 and are characterized by aγ -to-X-ray flux ratio in the range of∼260–780. None of them is either associated with a candidate Catalina counterpart or shows a significant X-ray time variability during the XMM–Newton ob-servation. We computed the 3σ detection limit by combining data from the pn, MOS1 and MOS2 detectors, as described in Baldi et al. (2002), corresponding to an unabsorbed X-ray flux of 9.6× 10−15erg cm−2s−1.
3.2.3 3FGL J1119.9−2204
The field of 3FGL J1119.9−2204 was observed by both Swift (67 ks) and XMM–Newton (73 ks). The analysis of the Swift observations was already reported by Hui et al. (2015), while here we focus on the unpublished XMM–Newton ones. The XMM–Newton observations (OBSID 0742930101) have been performed using the pn detector
Figure 4. WISE W1-band image (60 arcsec× 60 arcsec) of the candidate optical counterpart to 3FGL J0744.1−2523 (left) compared with the GROND g-band image (right). The candidate optical counterpart is marked by the two ticks. The mid-IR source matched in the AllWISE catalogue is at the centre of the image and is actually resolved in three or more stars in the GROND image, which has a finer pixel scale (0.158 against 2.75 arcsec) but covers the same area.
Figure 5. Observed colour–magnitude diagram for the stars detected by GROND in the 3FGL J0744.1−2523 field (dark blue points). The grey points correspond to the variable GROND source, with the intensity scale varying from the minimum (light grey) to the maximum of the light curve (dark grey). The dark and light cyan points correspond to a simulated stellar population from the Besanc¸on models (Robin et al.2004) for different values of the distance, that is,<10 kpc and 1–3 kpc, respectively.
in fast timing mode to search for pulsations in the brightest X-ray sources detected by Swift, whereas the MOS detectors were used in full-frame mode.
We combined the data from both MOS detectors to increase the sensitivity of the source detection, which we carried out as described in the previous sections. We found more sources close to the 3FGL error ellipse in the XMM–Newton MOS1+MOS2 observation with respect to the Swift observation (Hui et al.2015). The final source
list includes seven X-ray sources, with a combined detection signif-icance above 3.5σ . Fig.2shows the 0.3–10 keV exposure-corrected
XMM–Newton FoV obtained combining the images of both MOS
detectors. X-ray source properties are summarized in Table2. Three of these sources have a plausible Catalina counterpart, which are
Source #1, Source #2 and Source #6. None of these sources are
asso-ciated with a clearly periodic optical counterpart in the Catalina data.
Source #1 is coincident with the Swift source J1120_X1 of Hui et al.
(2015) and its candidate Catalina counterpart (J111958.3−220456; V= 15.6) was detected by both the Siding Springs Survey and
the Catalina Sky Survey in 234 and 76 observations, respectively. By running the Lomb–Scargle algorithm on the Catalina data, Hui et. al. (2015) identified a series of peaks in the periodogram that they considered to be caused by the non-uniform distribution of the data points. Our analysis of the Lomb–Scargle periodogram con-firms their conclusions. Source #1 is also associated with a GROND counterpart (g= 15.5). However, we did not detect any significant periodic flux modulation in the GROND data, whose cadence did not allow us to homogeneously sample the period space. This is true also for the GROND counterparts to Source #2 and Source #6. As a matter of fact, after inspecting the GROND images, we found that the latter is actually a galaxy.
Source #1 is the brightest of the X-ray sources detected within
the 3FGL error circle. Its unabsorbed X-ray flux in the 0.3–10 keV energy range isFX= 7.29+0.56−0.54× 10−14 erg cm−2 s−1, assuming,
as usual, a PL model with NH fixed to the integrated Galactic
value (NH = 4 × 1020 cm−2) and photon indexX= 2.63+0.12−0.11.
We detected neither a significant periodic modulation nor a clear short-term variability in Source #1, and this was true for the other
XMM–Newton sources detected in the 3FGL error ellipse. For the
3FGL J1119.9−2204 γ -ray flux Fγ = (1.68 ± 0.10) × 10−11 erg cm−2s−1(Acero et al.2015), the X-ray flux of Source #1 would yield Fγ/FX∼ 230. Source #1 was also the target of the XMM–
Newton/pn observations carried out in fast-timing mode. To perform
the periodicity analysis on Source #1, we used these data, which are characterized by a better timing resolution (0.03 ms) with respect to the MOS detectors (2.6 s), to search for X-ray pulsations and verify whether it was a suitable X-ray counterpart to 3FGL J1119.9−2204. Besides standard reduction and background screening, we cleaned
Table 2. Summary of the X-ray parameters of the sources detected within/close to the error ellipse of the Fermi-LAT source (see Fig.2), as discussed in the text. Here, we report the name of the 3FGL unassociated source, and for each X-ray plausible counterpart the best-fitting position, the best-fitting column density and photon index, the unabsorbed X-ray flux in the 0.3–10 keV energy band and theγ -to-X-ray flux ratio. All uncertainties are reported at the 68 per cent confidence level. For each observation, we also report the 3σ detection limits. In the last column, we flag the association with an optical counterpart to the X-ray source, detected by either Catalina or GROND, and flag whether it features a periodical modulation.
3FGL name X-ray J2000 X-ray coord. NH X Flux(0.3−10 keV) FFγ X
Detection limit Optical source RA, Dec (◦) (stat. err.a) (1022cm−2) (10−14erg cm−2s−1) (10−14erg cm−2s−1) counterparts
J0744.1−2523 – – – – – >525 4.54 YesP J0802.3−5610 1 120.7396,−56.1838 (0.92 arcsec) 0.18b 1.82+0.32 −0.30 4.93+1.46−1.16 264+82−67 0.96c No 2 120.5795,−56.2436 (1.15 arcsec) 2.17+0.48−0.44 2.60+1.20−0.80 500+235−161 No 3 120.5358,−56.2045 (1.65 arcsec) 2.08+0.61−0.51 2.66+1.40−0.77 489+261−148 No 4 120.6404,−56.1456 (1.56 arcsec) 1.39+0.54−0.49 3.10+1.59−1.31 419+219−181 No 5 120.6040,−56.0952 (1.20 arcsec) 4.08+0.74−0.68 9.62+5.72−3.63 135+81−52 YesP 6 120.5061,−56.1951 (1.59 arcsec) 2b 2.18+1.99 −1.74 596+547−479 No 7 120.5585,−56.2239 (1.98 arcsec) 2b 1.76+1.56 −1.35 739+658−571 No 8 120.4751,−56.0959 (1.98 arcsec) 2b 1.78+1.27 −1.12 730+525−464 No 9 120.4813,−56.1389 (1.95 arcsec) 2b 1.67+1.45 −1.28 778+680−601 No 10 120.4529,−56.2225 (1.41 arcsec) 2.40+0.50−0.45 2.50+1.34−0.75 520+283−163 No J1119.9−2204 1 169.9927,−22.0822 (0.38 arcsec) 0.04b 2.63+0.12 −0.11 7.29+0.56−0.54 230+22−22 0.41c Yes 2 170.0072,−22.0824 (0.72 arcsec) 1.99+0.18−0.17 1.97+0.35−0.32 853+160−148 Yesd 3 169.9970,−22.0543 (1.79 arcsec) 1.70+0.35−0.33 1.44+0.49−0.37 1167+403−308 No 4 170.0079,−22.0460 (1.43 arcsec) 2b 0.43+0.21 −0.20 3907+1922−1832 No 5 169.9702,−22.1102 (0.40 arcsec) 2.41+0.09−0.09 8.05+0.65−0.57 209+21−19 No 6 170.0036,−22.02489 (0.05 arcsec) 1.69+0.26−0.25 1.43+0.38−0.32 1175+320−272 Yese 7 169.9375,−22.0470 (1.92 arcsec) 2b 0.58+0.19 −0.22 2897+964−1112 No J1539.2−3324 – – – – – >1350 0.85 – J1625.1−0021 1 246.2935,−0.3578 (0.91 arcsec) 0.07b 3.19+0.26 −0.25 2.22+0.40−0.34 824+158−137 0.62c No 2 246.3162,−0.3296 (1.75 arcsec) 2b 0.64+0.40 −0.39 2859+1797−1752 Yese 3 246.2897,−0.3480 (0.85 arcsec) 1.53+0.21−0.20 2.94+0.73−0.57 622+160−127 Yes J2112.5−3044 1 318.1341,−30.7344 (1.03 arcsec) 0.07b 2.59+0.29 −0.28 1.41+0.39−0.22 1348+386−233 0.36c No 2 318.1060,−30.7171 (2.48 arcsec) 1.91+0.52−0.51 0.66+0.49−0.24 2879+2148−1068 No 3 318.1638,−30.7556 (1.07 arcsec) 2.16+0.51−0.45 0.79+0.45−0.21 2405+1381−663 No
aHere, we report only the 1σ statistical error, the 1σ systematic uncertainty is 1.5 arcsec for X-ray sources detected by XMM–Newton. bOwing to the low statistics in these sources, we fixed this parameter in the spectral analysis.
cHere, we computed the 3σ detection limit combining data from the pn, MOS1 and MOS2 detectors, as described in Baldi et al. (2002). dIdentified as a galaxy.
eTwo possible counterparts.
the pn data by the typical peculiar flaring background component produced by the interactions of charged particles with the CCD (see Burwitz et al.2004) following the same procedure as adopted by De Luca et al. (2005). We extracted the source photons using an 8 pixel wide strip (4.1 arcsec pixel size) containing 95 per cent of the source counts and the background photons from two stripes (4 and 4 pixels wide) away from the source region. All photon arrival times were converted to the Solar system Barycentric Dynamical Time with the
SAStask barycen and we used the FTOOL powspec to search for
periodic signal modulations. However, we did not detect any sig-nificant periodic flux modulation down to a period of∼1 ms. This, however, would only rule out the possibility that Source #1 is an iso-lated MSP, unless the pulsed fraction is too low for pulsations to be detected in the XMM–Newton observation or the X-ray emission is not pulsed because of an unfavourable viewing/beaming geometry. Unfortunately, the available X-ray data do not allow simultaneous fitting of both a spin and an orbital period. Therefore, we cannot firmly rule out that Source #1 is, indeed, an MSP in a binary system and, as such, the likely counterpart to 3FGL J1119.9−2204.
Other X-ray sources detected close to the 3FGL error ellipse are characterized by aγ -to-X-ray flux ratio 1000, except for Source
#5 (Fγ/FX∼ 200). The 3σ detection limit for the XMM–Newton
observation is 4.1× 10−15erg cm−2s−1, after combining data from both MOS detectors.
3.2.4 3FGL J1539.2−3324
Only Swift observations (84 ks total integration time) of the 3FGL J1539.2−3324 field have been obtained so far. We detected no candidate X-ray counterpart to the LAT source close to the 3FGL error box (see also Hui et al.2015). The computed 3σ sensitivity limit in the 0.3–10 keV energy range is 8.5× 10−15erg cm−2s−1. Itsγ -ray flux Fγ = (1.15 ± 0.10) × 10−11 erg cm−2s−1(Acero et al.2015) yieldsFγ/FX 1350 for 3FGL J1539.2−3324. We
observed the field of 3FGL J1539.2−3324 with GROND but no optically variable source has been detected in the 3FGL error el-lipse through an automated variability search.
Figure 6. Folded Catalina light curve of the star SSS J080225.1−560543 that is the optical counterpart of the XMM–Newton Source #5, possibly associated with 3FGL J0802.3−5610. The light curve has been folded at the best period of∼0.4159 d. A rebinning of 0.04 in phase has been applied for a better visualization.
3.2.5 3FGL J1625.1−0021
The 3FGL J1625.1−0021 field was observed by both Swift (9 ks) and XMM–Newton (26 ks). In the XMM–Newton observation (OBSID 0672990401), we found three sources close to the 3FGL er-ror ellipse (Fig.2). We identified the brightest X-ray source (Source
#1) with source J1625_X1 of Hui et al. (2015). Its unabsorbed X-ray flux in 0.3–10 keV energy range is FX= 2.22+0.40−0.34× 10−14 erg
cm−2s−1assuming a PL model with hydrogen column density fixed to the integrated Galactic value (NH= 5.8 × 1020cm−2) and photon
index X= 3.19+0.26−0.25. Alternatively, the X-ray spectrum can be
fitted by an absorbed BB with fixed NHand temperature kT= 0.18
± 0.02 keV. Our spectral analysis is in agreement with that of Hui et al. (2015). Source #2 is the faintest source, characterized by an unabsorbed X-ray flux FX= 6.4+0.40−0.39× 10−15 erg cm−2 s−1,
obtained by a PL model with fixed column density and photon index (X= 2). Source #3 is the brightest source detected closest
to the edge of the 3FGL error ellipse. Assuming a PL model with fixed column density andX= 1.53+0.21−0.20, the unabsorbed X-ray
flux is FX= 2.94+0.73−0.53× 10−14 erg cm−2 s−1. The two
bright-est X-ray sources, Source #1 and Source #3, would have an
Fγ/FX ∼825 and ∼622, respectively, assuming the
3FGL J1625.1−0021 γ -ray flux Fγ = (1.83 ± 0.12) × 10−11erg
cm−2s−1(Acero et al.2015), whereas theγ -to-X-ray flux ratio for the fainter X-ray source (Source #2) would be higher,∼2900. We did not detect either a significant periodic modulation or a clear short-term variability for the three plausible X-ray counterparts. The 3σ point source detection limit for the XMM–Newton observation is 6.2× 10−15erg cm−2s−1.
We detected plausible Catalina counterparts to Source #2 and Source #3, which are CSS J162516.0−001945/MLS J162515.8−001944 (V ∼ 20.07) and MLS J162509.5−002051 (V ∼ 22.06), respectively. However, neither of these two sources shows a significant periodic optical flux modulation. Regardless of an association with an X-ray source, we exploited the fact that the field is covered by the Catalina Surveys Periodic Variable Star Cat-alogue (Drake et al.2014) to search for variable optical sources. However, we found none. The closest variable Catalina source is CSS J162450.6−002135 (V = 16.48) with a period of 0.628 d, at a distance of 0.◦9 from the centre of the 3FGL error ellipse
(r95= 0.◦04). We could not find variable optical/IR sources to the three XMM–Newton sources in the GROND data either.
3.2.6 3FGL J2112.5−3044
The field of 3FGL J2112.5−3044 has been observed by both Swift (3.8 ks) and XMM–Newton (33.8 ks). In the XMM–Newton obser-vation (OBSID 0672990201), we detected three X-ray sources in the 3FGL error ellipse (see Fig.2). All spectra were fitted using a PL model with column density fixed to the integrated Galactic
NH, that is,∼6.9 × 1020 cm−2. Source #1 is the brightest X-ray
source, with an unabsorbed X-ray flux in the 0.3–10 keV energy range ofFX= 1.41+0.39−0.22× 10−14erg cm−2s−1andX= 2.59+0.29−0.28.
Source #2 and Source #3 are fainter, the PL model fitted with
fixed NH and X ∼1.91 and ∼2.16, respectively, yields an
un-absorbed X-ray flux of FX= 6.6+4.9−2.4× 10−15 erg cm−2 s−1 and
FX= 7.9+4.5−2.1× 10−15 erg cm−2 s−1. The Fγ/FX for three X-ray
sources would be between∼1300 and ∼3000 for Fγ = (1.90 ± 0.14)× 10−11 erg cm−2 s−1(Acero et al.2015). We did not de-tect either a significant periodic modulation or a clear short-term variability for the three possible X-ray counterparts. None of them is associated with a Catalina counterpart, whereas the field has not been observed by GROND. The computed 3σ detection limit of the XMM–Newton observation in the 0.3–10 keV energy range is 3.6× 10−15erg cm−2s−1.
3.2.7 Long-term variability
Since for most of our Fermi-LAT sources we have both Swift and
XMM–Newton observations, we used these data to search for
possi-ble long-term variability for the X-ray sources detected within/close to the 3FGL error ellipses, which might spot possible transitional RB candidates. The transition from a rotation-powered to an accretion-powered regime, due to the appearance of an accretion disc, yields a variation of the X-ray flux by a factor of∼10 (e.g. Bogdanov et al.
2015).
For each Swift observation, we extracted the source events from a circular region with radius of 25 arcsec centred on the position detected by the more sensitive instruments on board of XMM–
Newton, while to evaluate the background, we extracted events
from a source-free region of 130 arcsec radius. We generated the ancillary response files with the xrtmkarf task accounting for differ-ent extraction regions, vignetting and point-spread function correc-tions and we used the latest available spectral redistribution matrix (v014). We estimated the source flux by fitting in the 0.3–10 keV band to each spectrum a PL model, with hydrogen column density and X-ray photon index fixed to the best-fitting values obtained by
XMM–Newton, see Table2. In those cases in which the count rate in an observation was compatible with zero, we set an upper limit at the 3σ level.
Swift observed three times 3FGL J0802.3−5610 in 2012, 27 times
3FGL 1119.9−2204 from 2010 to 2013, twice 3FGL J1625.1−0021 from 2010 to 2011, and six times 3FGL J2112.5−3044 from 2013 to 2014. The flux values measured for each X-ray source with Swift and
XMM–Newton do not display any significant long-term variability.
Though almost all Swift observations provided 3σ upper limits on the flux because they are often very short (of the order of few ks) and the X-ray sources detected are very faint, these values are in agreement with those obtained by more sensitive instruments.
3.3 Candidates with a recent pulsar identification
3.3.1 3FGL J1035.7−6720
The field has been observed in X-rays both by Swift and XMM–
Newton (OBSID 0692830206) for a total integration time of 43 and
25 ks, respectively. A candidate X-ray counterpart to the LAT source has been found in the XMM–Newton data by Saz-Parkinson et al. (2016). The source X-ray spectrum is fit by a PL with photon index
X= 2.91+0.46−0.40, for a fixed NH= 2 × 1021cm−2, with an unabsorbed
0.3–10 keV flux FX= 3.06+0.97−0.50× 10−14 erg cm−2 s−1. For the
3FGL J1035.7−6720 γ -ray flux Fγ= (2.59 ± 0.14) × 10−11erg cm−2s−1(Acero et al.2015), the source X-ray flux corresponds to Fγ/FX∼ 850. The field of 3FGL J1035.7−6720 is not covered
by the Catalina Sky Survey but has been observed with GROND. No optical/IR counterpart, however, has been detected at the X-ray source position and no variable optical/IR object has been found in the rest of the 3FGL error ellipse through an automated variability search.
While this paper was close to being finalized, 3FGL J1035.7− 6720 was detected as aγ -ray pulsar by the Einstein@Home sur-vey (Clark et al.2017), although it is not neither known whether it is isolated or binary nor whether it is an MSP or a young pulsar. This source is under investigation by these authors be-cause of its peculiar timing properties and a more detailed ac-count will be given in a forthcoming publication (Clark et al. in preparation).
3.3.2 3FGL J1630.2+3733
The Fermi-LAT source 3FGL J1630.2+3733 has been recently identified by the Fermi Pulsar Search Consortium to be a binary MSP (PSR J1630+3734), with a spin period of Ps= 3.32 ms and
an orbital period Porb= 12.5 d (Ray et al.2012; Sanpa-Arsa et al.
in preparation). The pulsar coordinates are not published with a degree of accuracy good enough to identify its counterpart at other wavelengths based upon positional coincidence. None the less, we searched for a possible X-ray counterpart to PSR J1630+3734 in the
Swift data, which consists of only a snapshot 4.8-ks exposure. We
did not find X-ray sources close to the 95 per cent 3FGL error box. The computed 3σ sensitivity limit in the 0.3–10 keV energy range is 6.1× 10−14erg cm−2s−1. This represents the first constraint on the pulsar X-ray flux obtained so far. No observations with other X-ray satellites have been obtained for this source. Therefore, we can set a limit ofFγ/FX 110 for Fγ= (6.86 ± 0.10) × 10−12erg
cm−2s−1(Acero et al.2015).
Since the Swift observation is relatively shallow, the X-ray coun-terpart to the MSP might be below the detection limit. Therefore, like we did for the 3FGL J1625.1−0021 field, we searched for variable sources in the Catalina Surveys Periodic Variable Star Cat-alogue (Drake et al.2014) that might not be associated with a bright X-ray source. However, also in this case we found none. The field has been observed also by the Swift/UVOT but only six U-band exposures (4.8 ks total) are available, not enough to pinpoint a candidate companion star to the MSP based upon optical variabil-ity. Therefore, the companion star to PSR J1630+3734 remains unidentified.
3.3.3 3FGL J1744.1−7619
The field has been observed by both Swift and XMM–Newton (OBSID 0692830101) for a total integration time of 6.9 and 25.9 ks,
respectively (see also Hui et al.2015). A candidate X-ray counter-part to the LAT source has been found in the XMM–Newton data by Saz-Parkinson et al. (2016), identified as source J1744_X1 of Hui et al. (2015). The source X-ray spectrum is fit by a PL with photon indexX= 2.71+0.40−0.39, for a fixed NH= 8 × 1020 cm−2, with an
unabsorbed 0.3–10 keV fluxFX= 1.92+0.59−0.39× 10−14erg cm−2s−1
(Saz Parkinson et al.2016). For the 3FGL J1744.1−7619 γ -ray flux
Fγ = (2.25 ± 0.13) × 10−11erg cm−2s−1(Acero et al.2015), this corresponds to Fγ/FX∼ 1170. The field of 3FGL J1744.1−7619
is only covered by the Catalina Sky Survey, but no optical coun-terpart is found at the X-ray source position. Like in the case of 3FGL J1035.7−6720, 3FGL J1744.1−7619 has been recently de-tected as aγ -ray pulsar by the Einstein@Home survey (Clark et al.
2017) and it is also under investigation.
4 S U M M A RY A N D D I S C U S S I O N
As part of our survey of candidate MSPs, we identified a new candidate RB system (3FGL J2039.6−5618; Salvetti et al.2015) through correlated optical/X-ray flux modulations, we indepen-dently confirmed 3FGL J0523.3−2528 (Strader et al. 2014) and 3FGL J1653.6−0158 (Romani et al.2014) as binary MSP candi-dates, we found a new likely binary MSP candidate through the de-tection of a periodic optical modulation (3FGL J0744.1−2523). For 3FGL J0802.3−5610, we found only a marginal evidence of optical periodicity that may be affected by the presence of a strong daily pe-riodicity associated with the cadence of the Catalina observations. For this reason, follow-up optical observations are required to con-firm such periodicity and identify 3FGL J0802.3−5610 as a new binary MSP. For both 3FGL J1539.2−3324 and the recently identi-fied binary MSP 3FGL J1630.2+3733≡PSR J1630+3734 (Sanpa-Arsa et al. in preparation), we could find neither a candidate X-ray counterpart nor a variable optical counterpart candidate within the 3FGL error ellipse. For the other sources (3FGL J1035.7−6720, J1119.9−2204, J1625.1−0021, J1744.1−7619, J2112.5−3044), we detected up to several candidate X-ray counterparts but we could not find an association with a periodically modulated optical coun-terpart or with any optical councoun-terpart at all. Therefore, like for 3FGL J1539.2−3324, their proposed identification as binary MSPs cannot be confirmed yet, although both 3FGL J1035.7−6720 and 3FGL J1744.1−7619 have now been reported to be pulsars of still unspecified type (Clark et al.2017). For the others, deeper X-ray and optical observations are needed to ascertain their nature.
For 3FGL J2039.6−5618, we exploited archival WISE data (Mignani et al.2016) to obtain a better characterization of its multi-band spectrum with respect to Salvetti et al. (2015). Interestingly, its multiband UV-to-mid-IR spectrum shows a bump in the near/mid-IR emission, which can be explained by the presence of two BB com-ponents (Section 3.1.3). Since modulations are seen in the GROND optical bands (Salvetti et al. 2015), the hotter BB component is likely associated with the emission from the tidally distorted com-panion star. Since the same modulations are also seen in the JHK GROND light curves, the emission from the companion star must contribute in the near-IR as well. The colder BB, however, cannot be entirely associated with emission from the star, and its origin is unclear. As we suggested in Mignani et al. (2016), it might be asso-ciated with emission from cold intra/extra-binary material, perhaps associated with the ablated gas from the companion or a residual of an accretion disc, or both. Interestingly, the size of the emission region is a few times that of the star Roche lobe, which would corroborate the latter hypothesis. Setting tighter constraints on the absence of periodicity in the multi-epoch WISE data (Section 3.1.3)
